Вы находитесь на странице: 1из 25

The Society for Financial Studies

Signaling, Investment Opportunities, and Dividend Announcements


Author(s): Pyung Sig Yoon and Laura T. Starks
Source: The Review of Financial Studies, Vol. 8, No. 4 (Winter, 1995), pp. 995-1018
Published by: Oxford University Press. Sponsor: The Society for Financial Studies.
Stable URL: http://www.jstor.org/stable/2962297 .
Accessed: 05/12/2014 08:36
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Oxford University Press and The Society for Financial Studies are collaborating with JSTOR to digitize,
preserve and extend access to The Review of Financial Studies.

http://www.jstor.org

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment

Opportunities, and Dividend


Announcements
Pyung Sig Yoon
Laura T. Starks
University of Texas at Austin

This article examines potential explanationsfor


the wealth effects surrounding dividend change
We find that new information
announcements.
concerning managers' investment policies is not
revealed at the time of the dividend announcement. We also,find that dividend increases (decreases) are associated with subsequent signifin capital expendiicant increases (decreases)
tures over the threeyears following the dividend
change, and that dividend change announcements are associated with revisions in analysts'
forecasts of current earnings. These results are
consistent with the cash flow signaling hypothesis rather than the free cash flow hypothesis
as an explanation for the observed stock price
reactions to dividend change announcements.

It is widely accepted that announcements of changes


in dividend payouts affect firm value.1 There is an
ongoing debate, however, concerning why dividend

Previous drafts of this work were titled "Cash Flow Signaling Hypothesis
vs. Free Cash Flow Hypothesis: The Case of Dividend Announcements."
A previous version was presented at the annual meeting of the Western
Finance Association, Vancouver, Canada, June 1993. The authors would
like to thank Robert Bliss, David Chapman, Richard Green (the editor),
David Ikenberry, Meeta Kothare, Ken Lehn, John Martin, Roni Michaely,
Robert Parrino, A. J. Senchack, Tom Shively, two anonymous referees, and
especially Chris James and Paul Laux for helpful comments. Address correspondence to Laura T. Starks, Department of Finance, University of Texas
at Austin, Austin, TX 78712-1179.
Studies that document the wealth effects of dividend change announcements include Aharony and Swary (1980), Asquith and Mullins (1983), Bajaj and Vijh (1990), Eades, Hess, and Kim (1985), Kalay and Loewenstein
(1985, 1986), and Pettit (1972).
7he Review of Financial Studies Winter 1995 Vol. 8, No. 4, pp. 995-1018
1995 The Review of Financial Studies 0893-9454/95/$1.50

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

7he Review of Financial Studies / v 8 n 4 1995

changes affect firm value. At the heart of this debate is the question
of exactly what information is being conveyed to the market by the
dividend change. The primary explanation has been the cash flow signaling hypothesis as developed in theoretical models by Bhattacharya
(1979), John and Williams (1985), Kalay (1980), and Miller and Rock
(1985). These authors argue that since managers possess more information about the firm's cash flows than do individuals outside the
firm, the managers have incentives to unambiguously "signal" that
information to investors. According to these models then, dividend
changes convey managers' information about future and/or current
cash flows.
An alternate, although not mutually exclusive, explanation is that
changes in dividends reflect changes in managers' investment policies given their opportunity set [John and Lang (1991) and Lang and
Litzenberger (1989)]. This explanation is based on the free cash flow
hypothesis suggested by Jensen (1986). The free cash flow hypothesis asserts that managers with substantial free cash flow will invest
it at below the cost of capital or waste it on organizational inefficiencies rather than distribute it to shareholders. On the other hand,
these managers could change their investment policies and increase
dividends, thus paying out current cash that would otherwise be invested in low-return projects or wasted [Jensen (1986), p. 324]. While
Jensen (1986) does not explicitly state that changes in dividends reflect changes in the managers' investment policies, his free cash flow
hypothesis predicts that changes in wasteful investments for firms
with poor investment opportunities should have significant valuation
effects.
Lang and Litzenberger (1989) focus on this implication of the free
cash flow hypothesis. They argue that the free cash flow hypothesis does a better job of explaining stock price reaction to dividend
change announcements than does the cash flow signaling hypothesis.
According to their argument, a significant stock price reaction would
be observed for dividend changes when the dividends affect the level
of cash flows available for wasteful investments. The rationale is that
for firms that are overinvesting, a dividend increase implies a reduction in management's policy of overinvesting, while a dividend decrease implies further overinvestment. Thus, the information content
of a dividend change announcement depends on the severity of the
firm's agency problems, that is, how much the firm is overinvesting.
According to Lang and Litzenberger (1989) a significant stock price
response should be observed only when dividend changes affect investors' expectations about the size of the firm's future investment in
negative net present-value projects. Note then that this interpretation
of the free cash flow hypothesis requires that dividend change an-

996

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment Opportunities, and Dividend Announcements

nouncements provide information concerning changes in managers'


actions.
The difference between the cash flow signaling hypothesis and the
Lang-Litzenberger (1989) interpretation of the free cash flow hypothesis can be summarized as follows: Under the cash flow signaling
hypothesis, the dividend change provides information about current
and/or future cash flows, while under the free cash flow hypothesis, the dividend change provides information about changes in the
managers' misuse of cash flows.
Using Tobin's q ratio as a proxy for the overinvestment problem,
Lang and Litzenberger (1989) conclude that their empirical results are
more consistent with the free cash flow hypothesis than the cash flow
signaling hypothesis. They find a differential market reaction between
firms that are overinvesting and firms that are not. Based on analysts'
earnings forecasts, they also conclude that the dividend change does
not provide information about the firm's current cash flows.
A number of other studies have tested the free cash flow and cash
flow signaling hypotheses by examining whether the effects of the
firm's investment opportunities or the signaling of current and future cash flow is the primary explanation for stock price reaction to
firm announcements. Evidence has been mixed. While some empirical studies have presented evidence supporting the free cash flow
hypothesis,2 others have found evidence counter to its implications.3
Our interest is in determining which hypothesis is more appropriate for explaining the information conveyed in dividend change announcements. Accordingly, we analyze the extent to which the dividend changes themselves are related to the firms' investment opportunities and the extent to which the wealth effects from the dividend
change announcement are related to the investment opportunities or
cash flow signaling.
The results indicate that cross-sectional differences in observed dividend policy are related to investment opportunities. In a sample of
firms with large dividend changes we find that firms with poor investment opportunities have higher dividend yields. Such a relationship
is consistent with Jensen's (1986) free cash flow hypothesis in which

For example, evidence in favor of the free cash flow hypothesis is presented by Pilotte (1992) for
security offering announcements, Keown, Laux, and Martin (1992) for joint venture announcements, Lang, Stulz, and Walkling (1991) for tender offers, Perfect, Peterson, and Peterson (1994)
for self-tender offers, and Lehn and Poulsen (1989) for going-private transactions.

3 Servaes (1994) finds no evidence of overinvestment on the part of takeover targets. Howe, He, and

Kao (1992) find no difference between high-q and low-q firms in the market's reaction to one-time
cash flow events such as share repurchase and specially designated dividends. Denis, Denis, and
Sarin (1992) report that firms with Tobin's q less than one have significantly greater stock price
reactions to dividend change announcements largely because they pay higher dividends and their
dividend changes are of greater magnitude.

997

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

7he Review of Financial Studies / v 8 n 4 1995

firms with fewer growth opportunities should have higher dividend


yields. It is also consistent with Smith and Watts' (1992) empirical
finding that firms with more assets in place and fewer growth options
have higher payout ratios.
The focus in Jensen (1986) and Smith and Watts (1992) is on the
level of dividend payments, whereas our focus is on the change in
dividends. We find that the information revealed at the time of the
dividend change announcement is more consistent with the prediction
of the cash flow signaling hypothesis. After controlling for the size of
the dividend change, the anticipated dividend yield, and the market
value of the firm, there is no difference in the magnitude of stock
price reactions to dividend announcements across firms with different
investment opportunities (measured by Tobin's q ratio or an alternate
proxy, the direction of insider trading). This result is counter to the
prediction of Lang and Litzenberger's (1989) version of the free cash
flow hypothesis in which they contend that the absolute value of
the announcement abnormal return should be larger for firms with
poor investment opportunities than for firms with good investment
opportunities.
However, a regression analysis of the observed wealth effects segmented by investment opportunity sets cannot be considered definitive evidence against the free cash flow hypothesis because the control
variables (the magnitude of the change, the dividend yield, and the
market value of the firm) are also related to the firm's investment opportunities. A more appropriate approach to testing the free cash flow
and cash flow signaling hypotheses is to analyze the sources of the
wealth effects suggested by those two hypotheses. Consequently we
investigate (1) the extent to which dividend changes are related to
subsequent changes in wasteful investment (as predicted by the free
cash flow hypothesis) and (2) the extent to which dividend changes
are associated with changes in cash flow expectations (as predicted
by the cash flow signaling hypothesis).
If dividend changes reflect modifications in management's policy
toward overinvesting, then we should observe a change in capital expenditures following the dividend increase or decrease. Specifically
we should observe overinvesting firms reducing their capital expenditures after dividend increases. However, our results are not consistent
with this prediction of the free cash flow hypothesis. We find that, in
general, there are significant increases (decreases) in capital expenditures after dividend increases (decreases) for firms regardless of their
investment opportunities.
We also conduct a thorough investigation of whether changes in
investors' expectations associated with dividend change announce-

998

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment Opportunities, and Dividend Announcements

ments are related to current and/or future cash flow expectations.


Using revisions of analysts' forecasts for current earnings, we provide
evidence that announcements of dividend increases and decreases
cause analysts to revise their current earnings forecasts in a manner generally consistent with the cash flow signaling hypothesis. In
contrast, when we examine revisions of analysts' forecasts for future
earnings, we document significant changes only for dividend decreasing firms. Analysts do not significantly revise their expectations of future earnings for dividend increasing firms. This finding is particularly
important because previous researchers have not examined changes
in long-run earnings expectations around dividend change announcements, although it has been frequently stated that dividend changes release managers' information about both current and future cash flows.
This finding is also consistent with the empirical evidence that a dividend decrease results in a larger stock market reaction than would
an equivalent dividend increase.
In the next section we describe the data and explain our proxies
for the quality of the investment opportunity set. We then analyze firm
characteristics between firms with good investment opportunities versus firms with poor investment opportunities in Section 2. After we
examine the wealth effects of dividend change announcements in Section 3, we investigate the sources of these wealth effects by analyzing
changes in capital expenditures in Section 4 and changes in cash flow
expectations in Section 5. Concluding comments are presented in Section 6.
1. Data and Proxies for Investment

Opportunity

Set

1.1 Data
The sample of 3748 dividend increase and 431 dividend decrease
announcements over the period 1969 to 1988 consists of all NYSE
stocks from the Center for Research in Security Prices (CRSP) monthly
master file that satisfy the following criteria:
1. We restrict our attention to regular quarterly U.S. cash dividends
per share. Based on the naive expectations model, the unexpected
dividend change is defined to be the proportional change in dividends from the previous quarter. Warther (1994) argues that due to
the coarseness of the signaling equilibrium, not all dividend changes
contain information. In order to ensure that any potential information signal is significant, we impose the restriction that the dividend
change must be at least 10 percent.
2. The announcement does not represent a dividend initiation or
omission.

999

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

The Review of Financial

Studies/

v 8 n 4 1995

3. A stock split or stock dividend does not occur during the month
before or the month in which the dividend change announcement is
made.
4. Daily return data for the 200 trading days surrounding the announcement are available from the CRSP daily return master file.
5. Empirical estimates of Tobin's q ratios are available from the
National Bureau of Economic Research (NBER) manufacturing sector
master file.4
1.2 Proxy for the investment opportunity set
A central issue in any test of the free cash flow hypothesis is the
question of a measure for firms' investment opportunities. We employ
two proxies, the often-used Tobin's q ratio and the direction of insider
trading. As suggested by Lang and Litzenberger (1989) and Lang, Stulz,
and Walkling (1991) our first proxy for the investment opportunities
available is Tobin's q ratio, defined as the ratio of the market value of
the firm's assets to their replacement costs. Although most empirical
estimates of Tobin's q ratios are built on Lindenberg and Ross (1981),5
the classification of q ratios into high-q and low-q firms varies by
author. For example, Lang and Litzenberger define a value-maximizing
firm as one with a one year q greater than unity, while Lang, Stulz,
and Walkling (1991) define a high-q firm as one with a three year
average q greater than one. Since a cutoff of one has some theoretical
appeal, we take a one-year q greater than one as a basic cutoff point
for high- and low-q firms. We also use two other cutoff points: a three
year average q greater than unity and a one year q greater than the
median within a calendar year.6
Although the estimated Tobin's q ratio is commonly employed as
a proxy for the investment opportunity set, it has several potential
problems. First, the estimate is of the average q ratio, but as pointed

the NBER file contains only industrial firms, the problems associated with dividend announcements for regulated firms are avoided.

4Since

5 The empirical estimates of Tobin's q ratios are computed slightly differently by authors, but most of

them are built on Lindenberg and Ross (1981). Perfect and Wiles (1994) construct four procedures
to estimate q ratios (primarily based on the methodology developed by Lindenberg and Ross).
Their results indicate that the methods tend to produce equivalent empirical results with one
exception: a q ratio computed using book values of long-term debt and total assets.
6 Servaes (1991) provides reasons why a cutoff of one may not be appropriate. For example, the

median Tobin's q may differ from one. Specifically, examining the median Tobin's q across all
firms in the NBER file each year over the period 1968 to 1987, we find that the median ranges
from 0.61 (in 1975) to 1.93 (in 1969). The median Tobin's q by year for our sample is consistent
with these population parameters. Since we find that the equality of the medians across years
can be rejected at any reasonable significance level, we checked the robustness of our results by
using the median Tobin's q ratio by year as an alternative cutoff point to separate high-q firms
from low-q firms. We also test the results with an industry median q ratio as a cutoff point. Our
results are not sensitive to the q classification process.

1000

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling,

Investment

Opportunities,

and Dividend

Announcements

out by Lang and Litzenberger (1989), the differences in investment


opportunities would be found in the marginal q ratio. Second, measurement errors induce noise in estimated q ratios. Third, estimated q
ratios may be based on outdated information. In calculating Tobin's q
ratio, the market value of the firm's assets and their replacement cost
are evaluated at the year end before the announcement. For example,
if one assumed that dividend announcements are made evenly across
the following year, the estimated q ratios would be six months old, on
average, and consequently would not contain any private information
developed just prior to the dividend change announcements.
Because of these problems, we also employ a second proxy for
the investment opportunities available. John and Lang (1991) suggest that the direction of insider trading activity combined with the
dividend announcements may be a better proxy for the investment
opportunity set than the average q ratio. Following John and Lang,
we construct an insider trading index based on insider open market
sale and purchase transactions for the period of two quarters prior
to the dividend change announcements. Since our results using this
alternate proxy are consistent with the Tobin's q results, we do not
report them here.
2. Differences
Firms

in Firm Characteristics

Between High-q and Low-q

We begin our empirical analysis by comparing firm characteristics of


high-q firms with those of low-q firms in order to identify any significant differences between the two groups. Previous studies have
identified three factors as influential on the stock market reaction to
dividend change announcements: the firm's dividend yield, the firm's
size, and the magnitude of the dividend change. The first variable,
dividend yield, has been suggested as a proxy for clientele effects.
Bajaj and Vijh (1990) posit that if investors with preferences for dividends are the marginal investors in high-yield stocks, the price reaction to a dividend change should be larger, the higher the anticipated
yield of the stock. They find that for high-yield stocks, price reactions
to dividend increases are significantly more positive and to dividend
decreases significantly more negative. Fehrs, Benesh, and Peterson
(1988) report similar results. The second variable reflects omitted market pricing factors such as the information asymmetry between large
versus small firms. Eddy and Seifert (1988) document that abnormal returns from the announcements of large dividend increases are greater
for small firms than for large firms. Finally, the third variable is suggested by dividend signaling models which predict that large dividend
changes are used to signal large changes in cash flows.

1001

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

7Te Review of Financial Studies / v 8 n 4 1995

The free cash flow hypothesis predicts that these firm characteristics should systematically differ between high-q versus low-q firms.
In particular, Jensen (1986) argues that firms with more growth opportunities should have lower dividend yields. Consistent with this
prediction, Smith and Watts (1992) document that firms with more
assets-in-place and fewer growth options have higher dividend payout ratios. [This finding is also consistent with Easterbrook (1984) and
Rozeff (1982).] In addition, a firm's size is related to its q. As pointed
out by Smith and Watts, size is a function of the firm's investment
opportunity set. That is, those firms with more growth opportunities
are more likely to become larger. Thus, the free cash flow hypothesis
based on contracting arguments predicts a negative relationship between dividend yields and Tobin's q ratios, and a positive relationship
between size and Tobin's q.
On the other hand, cash flow signaling models generally do not
have direct predictions for differences across high-q and low-q firms
in terms of dividend yield, firm size, and the magnitude of dividend
changes.
Panel A in Table 1 reports the averages of dividend change, dividend yield, and firm size by the sign of the dividend changes and
the level of the Tobin's q ratios.7 In our sample there are 3748 dividend increases and 431 dividend decreases announced from 1969
through 1988. While there is not a large difference in the proportion
of dividend increase announcements for high-q and low-q firms, for
dividend decreases, 87 announcements (20 percent) are classified to
a group of high-q firms and 344 (80 percent) to a group of low-q
firms. This result is consistent with the view that firms in general cut
their dividend payments when their performance is poor [DeAngelo,
DeAngelo, and Skinner (1992)].
We find that the characteristics of dividend change announcements
for high-q firms tend to be systematically different from those of lowq firms. For both dividend increases and decreases, the means of the
anticipated dividend yield and the dividend change for high-q firms
are significantly smaller than those for low-q firms. In addition, for
dividend increases there is a significant positive relationship between
firm size and Tobin's q ratios: the equality of mean firm size between
the two groups can be rejected at any reasonable significance level.

I The anticipated dividend yield in this article is measured by dividing the sum of all dividend

payments for the year preceding the announcement by the end-of-year stock price. The firm's
size is measured as the market value of the firm's assets at the end of the year preceding the
announcement. The dividend change is computed by dividing the dividend change in dollars by
the end-of-month stock price before the announcement.

1002

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment Opportunities, and Dividend Announcements

Table 1
Dividend

change, yield, and firm size by sign of dividend

change and Tobin's q'

Panel A: Characteristics of dividend change announcements


No. of observations
Average change
A.1. Dividend increases
q < 1
q> 1

2062 (55%)
1686 (45%)

Average yield

1215
2054

0.042
0.022

0.0021
0.0012

Mean difference
(q < 1) - (q > 1)
A.2. Dividend decreases
q < 1
q > 1

Average size

0.0009
(25.96)
344 (80%)
87 (20%)

-0.0082
-0.0061

Mean difference
(q < 1) - (q > 1)

-839
(-16.32)2

0.020
(41.68)

1,015
1,194

-0.0021
(-3.20)

0.072
0.040

-179
(-0.24)2

0.032
(12.52)

Panel B: Spearman correlation matrix for all sample3


Change

Log(size)

Yield

Tobin's q

1.00
-0.18
0.24
-0.37

1.00
-0.17
0.29

1.00
-0.65

1.00

Change
Log(size)
Yield
Tobin's q

Our sample consists of 3748 dividend increases and 431 dividend decreases announced over the
period 1969 to 1988 that meet the following criteria: (1) The announcement date is available
from the CRSP monthly master file. (2) Daily return data for the 200 trading days surrounding the
announcements are available. (3) The announcement does not represent a dividend initiation or
omission. (4) A stock split or stock dividend does not fall a month before or during the month
in which the announcement is made. (5) The dividend change is at least 10 percent compared
with the previous quarter. (6) Empirical estimates of Tobin's q ratios are available from the NBER
manufacturing sector master file. Tobin's q ratios are estimated as the market value of the firm's
assets divided by replacement costs, both are evaluated at the year end before the announcement
from the NBER tape. The change is computed by dividing dividend change in dollars by the
end-of-month stock price before the announcement. The size is the year-end market value of the
firm (in million $) obtained from the NBER tape. The yield is measured by dividing all dividend
payments for a year before the announcement by the end-of-year stock price. T-statistics are in
parentheses.
2

T-statistics are based on the logarithm of firm size.

3 All correlations are significant at the 0.01% level.

However, for dividend decreases, the mean differences in firm size


between the two groups are not significant.
The evidence to this point indicates that there are cross-sectional
differences in dividend changes that are related to a firm's investment
opportunities: low-q firms have higher dividend yields and larger dividend changes, and are smaller in size. These relationships indicate
a potential problem in testing the free cash flow hypothesis directly
against the cash flow signaling hypothesis. For example, while dividend yield reflects dividend clientele effects, it could also be regarded
as another proxy for investment opportunities. In the extreme case,
firms with many good investment opportunities would pay no dividends at all. This point is further reflected in the Spearman rank

1003

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

7he Review of Financial Studies/ v 8 n 4 1995

correlations between the Tobin's q ratios and the three control variables presented in Panel B of Table 1. The correlations between the
q ratio and dividend yield, dividend change, and firm size are all statistically significant, being -0.65, -0.37, and 0.29, respectively, suggesting that the control variables may also be proxying for investment
opportunities.
In this section we have pointed out that the dividend change itself
depends on a firm's investment opportunities. In the next section we
measure the stock market reaction to the dividend change announcement, conditional on the market's prior information concerning the
firm's investment opportunities.
3. Investment

Opportunities

and Abnormal Returns

The free cash flow hypothesis implies that dividend changes by low-q
firms will lessen or aggravate the overinvestment and accordingly affect the market value of the firm. On the other hand, dividend changes
by high-q firms would not be expected to have a particular effect on
stock prices. There is no reason that a change in the dividend should
affect the level of the optimal investment because the firms are not assumed to be overinvesting [see Lang and Litzenberger (1989)]. Thus,
according to their hypothesis, dividend change announcements for
high-q firms should not have information content, not because they
are fully anticipated, but because dividend changes do not affect the
market's assessment of managers' investment policies. (Under some
plausible assumptions, Lang and Litzenberger derive these predictions
in their model.) Thus, for either dividend increases or decreases, the
prediction is that announcements of dividend changes by high-q firms
should have no significant impact on the firms' stock prices, whereas
announcements of dividend changes by low-q firms should have a
substantial effect on stock prices, implying that the absolute value of
the abnormal return is larger for low-q firms than for high-q firms. In
contrast, the cash flow signaling hypothesis predicts significant price
reactions regardless of the level of q ratios, implying a symmetrical
impact between high-q and low-q firms for either dividend increases
or decreases.
Table 2 reports three day cumulative average abnormal returns
(CAAR), summed over days -1 to + 1 relative to the announcement
date, and average abnormal returns on the announcement day by the
sign of the dividend changes and by the level of the Tobin's q ratios. Abnormal returns are estimated from the market model using the
CRSP equally weighted index and Scholes-Williams betas.8 We find
8 The estimation period covers days -100

to -8 and days +8 to +100 (a total of 186 days).

1004

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment Opportunities, and Dividend Announcements

significant cumulative average abnormal returns for all four groups,


ranging in absolute value from 0.67 percent for dividend increases of
high-q firms to 5.30 percent for dividend decreases of low-q firms.
Note that the free cash flow hypothesis implies that for high-q firms,
dividend changes should have no impact on the managers' investment
policies, and therefore no impact on firms' stock prices. The significance of abnormal returns for all four groups is consistent with the
predictions of the cash flow signaling hypothesis.9 The results for the
average abnormal returns on the announcement day are consistent
with those for the cumulative average abnormal returns.
We find a larger share price reaction for dividend decreases than for
increases. This result is consistent with the cash flow signaling models
of Bhattacharya (1979) and Kalay (1980). Bhattacharya assumes that
the cost of making up a cash flow deficit is more than the benefit of a
cash flow surplus of the same size. Kalay argues that managers' reluctance to cut dividends is a necessary condition for dividends to convey information. Thus, both papers suggest that dividend decreases
are more costly. On the other hand, as posited by Lang and Litzenberger (1989), the free cash flow hypothesis implies that for high-q
firms, neither dividend increase nor decrease announcements should
have any impact on stock prices.
For dividend decreases, the cumulative average abnormal return
for high-q firms is not significantly different from that of low-q firms;
a result counter to the free cash flow hypothesis. In contrast, for dividend increases, the cumulative average abnormal return for low-q
firms is significantly larger than that of high-q firms. This result is not

The announcement period abnormal returns are summed over days -1 to +1 due to possible
information leakage and announcements being made after trading hours on the announcement
day. Our results remain qualitatively the same with the CRSP value-weighted index rather than
the CRSP equal-weighted index as our market proxy. Similar results are also obtained when we
calculate the two day cumulative abnormal returns.
9 While, on average, our results indicate that stock prices react favorably to dividend increases and

unfavorably to dividend decreases, these results are not uniform across the sample. The stock
price reactions to dividend increase announcements for 43 percent of the high-q firms and 34
percent of the low-q firms are negative. Similarly, 24 percent of the high-q firms and 18 percent
of the low- firms have positive stock price reactions to dividend decrease announcements. While
these results could be interpreted as evidence counter to the cash flow signaling hypothesis, they
may be due to the problem of determining the market's conditional expected dividend, especially
for dividend increases. In our empirical analysis, we have adopted the naive dividend expectation
model, which implies that the expected dividend change is zero on average. However, this model
may not be realistic, not only because many firms tend to increase dividends in the same quarter
every year, but also because the model does not incorporate the market's most recent expectation
since the last dividend payment. The lower percentage of stock price reactions in the opposite
direction for dividend decreases than dividend increases supports the view that the problem of
determining the market's conditional expected dividend is more serious for dividend increases.
Regardless, the fact that there are a number of reactions in the opposite direction to the dividend
change does not present a serious problem in testing which hypothesis is more consistent with
the observed market reactions because the competing free cash flow hypothesis also does not
predict negative reactions to dividend increases and positive reactions to dividend decreases.

1005

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

The Review of Financial Studies / v 8 n 4 1995

Table 2
Cumulative

average abnormal

returns by sign of dividend

Dividend
increases
q<1
No. of obs.
CAAR
Percent positive
AAR

changes -and level of q ratios'


Differences in absolute
values of means
(decreases - increases)

Dividend
decreases

2062
1.537 (17.41)
66%
0.969 (16.70)

q> 1
No. of obs.
CAAR
Percentage positive
AAR

1686
0.670 (7.18)
57%
0.350 (6.13)

(q < d)- (q > 1)


Mean difference on CAAR
Mean difference on AAR

0.867 (6.75)
0.619 (7.61)

-5.299
-3.197

344
(-14.67)
18%
(-12.24)

-2.689

87
(-6.14)
24%
(-5.13)

-0.700
-0.508

(-0.86)
(-0.87)

-4.599

3.762 (10.11)
2.228 (8.33)

3.929 (5.21)
2.339 (4.44)

Our sample consists of 3748 dividend increases and 431 dividend decreases announced over the
period 1969 to 1988 that meet the following criteria: (1) The announcement date is available
from the CRSP monthly master file. (2) Daily return data for the 200 trading days surrounding the
announcements are available. (3) The announcement does not represent a dividend initiation or
omission. (4) A stock split or stock dividend does not fall a month before or during the month
in which the announcement is made. (5) The dividend change is at least 10 percent compared
with the previous quarter. (6) Empirical -estimates of Tobin's q ratios are available from the NBER
manufacturing sector master file. Tobin's q are estimated as the market value of the firm's assets
divided by replacement costs, both are-evaluated year end before the announcement from the
NBER tapes. Abnormal returns are estimated from the market model using the CRSP equally
weighted index and Scholes-Williams betas. The estimation period is over days -100 to -8 and
days 8 to 100. CAAR refers to the three day (day -1, day 0, and day 1) cumulative average
abnormal returns. AAR refers to the average abnormal returns on the announcement day. Percent
positive refers to the percentage of positive cumulative abnormal returns. T-statistics are reported
in parentheses.

predicted by the cash flow signaling hypothesis, but is predicted by


the free cash flow hypothesis.
Previous articles document that the abnormal returns to dividend
increase announcements are positively related to dividend change and
dividend yield, but negatively related to the firm size. We have shown
in Table 1 that for dividend increases, low-q firms have higher dividend change, higher dividend yield, and smaller firm size than high-q
firms. Accordingly, we include these three control variables along with
a dummy for high-q firms, Dhigh q, in a regression of the cumulative
abnormal returns of the dividend increase announcements.
CAR=

0.50
(t = 1.51)

481 CHANGE
(t = 8.73)

19.08YIELD
(t = 4.50)

0.14 LOG(SIZE) +
(t = -3.28)

0.09 Qigh q*
(t = 0.57)

1006

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment Opportunities, and Dividend Announcements

F-statistic = 4.75,

R2 = 0.05.

The regression indicates that the relationships between the abnormal


returns and each of the three control variables are consistent with the
findings of previous studies, and that Tobin's q ratios are not significantly related to abnormal returns once the three control variables are
considered.
These results indicate that a differential price reaction between
high-q and low-q firms for dividend increases may be caused by
a difference in firm characteristics rather than differences in investment opportunities. Our regression results are not definitive evidence
against the free cash flow hypothesis because dividend change, dividend yield, and firm size are all related to a firm's investment opportunity. Accordingly, in the next section we directly examine the
sources of the wealth effects suggested by the two hypotheses.
4. The Effect of Dividend Change Announcements
Expenditures

on Capital

We have presented evidence that the information revealed in the dividend change announcement appears to be more of a reflection of
cash flows than a reaction to managers' actions in the context of
the free cash flow hypothesis. A further test of this conjecture can
be obtained by examining whether changes in the firm's investment
policies after the announcements are consistent with the free cash
flow hypothesis predictions. Under the free cash flow hypothesis, an
announcement of a dividend change will have an effect on the firm's
stock price if the size of the firm's future investment in negative net
present value projects is expected to change. Thus, since low-q firms
invest in negative net present value projects, there should be a decrease in wasteful capital expenditures after dividend increases and an
increase in wasteful investment after dividend decreases. The implication is that a change in capital expenditures would cause significant
stock price reactions for low-q firms. It also predicts no significant
change in capital expenditures for high-q firms, implying no effect on
stock prices.
We use capital expenditures (COMPUSTATdata item 128) to measure new investment by the dividend change firms. The analysis measures the percentage changes in capital expenditures in the first three
full fiscal years after a dividend change (years +1, +2, and +3) compared to the last fiscal year before a dividend change (year -1).
We measure capital expenditures in levels and as a fraction of the
end-of-period total assets. To control for industry effects, we also
present an industry-adjusted percentage change in capital expendi-

1007

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

7he Review of Financial Studies / v 8 n 4 1995

Table 3
Relations
tures,

between

dividend

change

Panel A: Dividend increases and q

<

announcements

and subsequent

capital expendi-

1
From year i to year j
-1 to +1

-1 to +2

-1 to +3

n = 1626

n = 1626

n = 1517

Percentagechange

39.68%***

52.10%***

Industry adjusted percentage change

8,31%***

1.13%***

65.87%***
1,11%***

n = 1626

n = 1626

n = 1517
11,13%***
0.70%***

A.1. Capital expenditures

A,2. Capital expenditures/Assets


Percentagechange

12.22%***

11.54%***

Industry adjusted percentage change

4,59%***

0.30%***

Panel B: Dividend increases and q > 1


From year i to year j
-1 to +1

-1 to +2

-1 to +3

n = 1168

n = 1175

n = 1083

Percentagechange

41.63%***

Industry adjusted percentage change

9.82%***

51.05%***
0.76%***

69.80%***
1.37%***

n = 1168

n = 1175

n = 1083

6.38%***
5.25%***

0.70%
0.00%

-0.57%
0.00%

B.1. Capital expenditures

B.2. Capital expenditures/Assets


Percentage change
Industry adjusted percentage change

Significance at the 5 percent level.


Significance at the 1 percent level.
'Median percentage change and industry adjusted change in capital expenditures and in capital
expenditures as a percentage of assets by sign of dividend changes and q ratios. Year-1 is the fiscal
year ending prior to dividend change announcements. Year +1 is the first full fiscal year after
dividend change announcements. Significance levels are based on two-tailed Wilcoxon signed
rank tests. Industry adjusted change for a given period equals the difference between the change
for dividend change company and the median change for a sample of companies in the same
industry during that period. The firms in the same industry are those that have the same four-digit
SIC code. The observation was excluded if there are less than three firms in the same industry.

tures. The industry-adjusted percentage change is defined as the percentage change in capital expenditures minus the median percentage
change over the same period for all firms in the same four-digit SIC
code as the dividend change firm.10 The results by sign of dividend
changes and level of q ratio are reported in Table 3.
Table 3 shows that there are significant increases (decreases) in
the level of capital expenditures after dividend increases (decreases).
This result holds regardless of the industry adjustment. The change in
capital expenditures is more pronounced for low-q firms than highWhen there were fewer than three firms in the same industry, that observation was excluded.
The results remain the same under the "same industry" definition with the two-digit SIC code.

1008

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment Opportunities, and Dividend Announcements

Table 3
(Continued)
Panel C: Dividend decreases and q

<

1
From year i to year j

C.1. Capital expenditures


Percentage change
Industry adjusted percentage change

C.2. Capital expenditures/Assets

-1 to +1

-1 to +2

-1 to +3

n = 269

n = 266

n = 245

-37.76%***
i24.92%***

- 18.49%***
-19.71***

-3.70%
-15.51%***

n = 269

n = 266

n = 245

-10.50%***
-2.76%***

Percentagechange

-33.39%***

-20.38%***

Industry adjusted percentage change

-16.30%***

-8.16%***

Panel D: Dividend decreases and q > 1


From year i to year j

D1. Capital expenditures


Percentage change
Industry adjusted percentage rate

D.2. Capital expenditures/Assets


Percentage change
Industry adjusted percentage change

-1 to +1

-1 to +2

-1 to +3

n = 67

n = 72

n = 68

1.29%
-14.53%***

-4.53%
-17.32%***

16.26%**
-13.19%

n = 67

n = 72

n = 68

-13.47%
-3.19%

-21.00%**
-5.35%***

-9.63%**
0.00%

** Significant at the 5 percent level.

Significant at the 1 percent level.

q firms for both dividend increases and decreases. This is in sharp


contrast to the implication of the free cash flow hypothesis that dividend increases would reduce overinvestment and dividend decreases
would increase wasteful investment.11 For low-q firms with dividend
increases (shown in Panel A), the level of investment significantly
increases in years +1, +2, and +3 relative to year -1. The industryadjusted changes are also significant for all periods, although they
are small in years +2 and +3. For completeness, Panel B contains
the changes in capital expenditures for high-q firms with dividend increases. These results are similar to those for the low-q firms. Panels C
and D indicate that for both high- and low-q firms that decrease their
dividends, there are reductions in capital expenditures over the next
three years. In particular, for the low-q firms, the industry-adjusted
changes are all negative and significant, equal to -24.92 percent,
-19.71 percent, and -15.51 percent.

If the firm issues debt or equity to finance a dividend increase, a change in dividends will not
induce a corresponding change in investment. The conclusions are not affected by the exclusion
of those cases from our sample.

1009

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

The Review of Financial Studies / v 8 n 4 1995

The analysis provided in this section directly examines the source


of the valuation effects suggested by the free cash flow hypothesis.
We find no evidence that dividend increase (decrease) firms reduce
(increase) their level of investment.12 On the other hand, our finding
is not inconsistent with the cash flow signaling hypothesis; if dividend
changes signal management's belief about the firm's future prospects,
dividend increase firms are able to invest more and dividend decrease
firms could be expected to cut capital expenditures.
5. Dividend Change Announcements
Earnings Expectations

and Current versus Future

Theories and empirical tests of the cash flow signaling model as an


explanation for stock market reaction to dividend announcements typically state that dividend announcements provide information about
current and/or future cash flows. According to Miller and Rock (1985,
p. 1037), dividend announcements act as "the missing piece of the
sources/uses constraint which the market needs to establish the firm's
current earnings." According to their model, although dividend announcements may provide information about future expected earnings, it is only indirectly. In this section we investigate the extent to
which the dividend announcement provides information about current and future earnings expectations. To do so, rather than inferring
these changes from observed stock price reactions which are also
affected by other factors, we use a direct measure of changes in investors' expectations: changes in analysts' earnings forecasts.
5.1 Analysts' earnings forecasts as a proxy for cash flow
expectations
There is mixed evidence concerning whether analysts change their
forecasts of current earnings after dividend announcements. Ofer and
Siegel (1987) report that analysts revise their earnings forecasts fol12

The relationship between the previously observed wealth effects and the investment opportunity
set depends to some extent on how we define high- and low-q firms. For example, if we happen
to assign low-q firms to the high-q firm category by mistake, we could erroneously observe a
significant stock price reaction to dividend change announcements of high-q firms. In contrast,
our results employing capital expenditure changes are not diminished by a poor proxy. The free
cash flow hypothesis predicts a significant change in wasteful capital expenditures in the direction
opposite to that of dividend changes for low-q firms. It also predicts no particular change in capital
expenditures for high-q firms. In this case, if we assign low-q firms into the high-q firm category,
we may observe a significant decrease in capital expenditures after dividend increases and a
significant increase after dividend decreases for both high- and low-q firms. Conversely, if we
assign high-q firms into the low-q firm category, we may observe an insignificant change in capital
expenditures for both high- and low-q firms. Our evidence that dividend increase (decrease) firms
increase (reduce) their level of investment significantly for both high- and low-q firms is not related
to how we determine high- and low-q firms. This is strong evidence against the free cash flow
hypothesis as an explanation for the wealth effects of dividend change announcements.

1010

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment Opportunities, and Dividend Announcements

lowing the announcement of an unexpected dividend change. They


find that the revision is positively related to the size of the unexpected dividend change. Similarly, Healy and Palepu (1987) show
that dividend initiations and omissions are leading indicators of superior and inferior earnings performance, respectively. On the other
hand, Lang and Litzenberger (1989) report that announcements of
sizable dividend changes are not significantly related to changes in
analysts' forecasts. In this section we analyze the extent to which the
announcement of an unexpected dividend change has a significant effect on the revision of current and future cash flow expectations. The
proxies for investors' cash flow expectations are analysts' forecasts for
the current year earnings and their five year growth forecasts. These
forecasts are obtained from the Institutional Broker Estimate System
(IBES) database developed by IBES Inc. The IBES database used here
contains summary statistics of analysts' earnings forecasts made by
major brokerage firms for about 2000 firms listed on the NYSE and
AMEX over the period 1976 to 1989. Since the IBES database shortterm analysts' earnings forecasts covers the period 1976 to 1989, our
sample is reduced to 2505 dividend increases and 205 dividend decreases. Similarly, the forecasts of the five year earnings growth rates
are available from 1981 to 1989, so that the sample is further reduced
to 883 increases and 131 decreases.
5.2 Revisions of short-term analysts' earnings forecasts
The cash flow signaling hypothesis predicts that dividend change announcements will cause investors to revise their cash flow expectations in the same direction as the dividend surprise. To test this
implication we measure the elasticity of the change in the median of
analysts' current earnings forecasts with respect to dividend changes,
computed by dividing the percentage change of the postannouncement median earnings forecast (compared to the preannouncement
median earnings forecast) by the percentage dividend change. The
use of the median provides more conservative statistics than the mean
because the median is less sensitive to extreme earnings revisions.
Table 4 reports the mean elasticity of the change in the median
of analysts' forecasts with respect to dividend changes. Similar to the
Healy and Palepu (1987) and Ofer and Siegel (1987) results, and in
contrast to the Lang and Litzenberger (1989) results, we find that both
dividend increases and decreases are associated with positive mean
elasticity. In addition, consistent with the observed larger stock price
reaction to dividend decreases, the magnitude of forecast revisions
is greater for the dividend decreases than the increases. We also divide the dividend increases and decreases into two groups according
to their investment opportunities (i.e., Tobin's q ratio). Three of the

1011

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

The Review of Financial Studies / v 8 n 4 1995

Table 4
The impact of dividend

change announcements
q <

on current cash flow expectations'


q
q>

1(q

< l)(q>1

Panel A: Dividend increases


No. of observations
Average elasticity
(t-stat p-value)
Positive/0/negative

1221

1284
0.08
(0.00)
636/313/335

(0.06)
468/400/353

171
0.56
(0.00)
145/14/12

34
0.27
(0.00)
22/7/5

0.01

0.07
(0.00)

Panel B: Dividend decreases


No. of observations
Average elasticity
(t-stat p-value)
Positive/0/negative

0.29
(0.00)

Our sample consists of 2505 dividend increases and 205 dividend decreases announced over the
period 1976 to 1988 that meet the following criteria: (1) The announcement date is available
from the CRSP monthly master file. (2) Daily return data for the 200 trading days surrounding
the announcements are available. (3) The announcement does not represent a dividend initiation
or omission. (4) A stock split or stock dividend does not fall a month before or during the
month in which the announcement is made. (5) The dividend change is at least 10 percent
compared with the previous quarter. (6) Empirical estimates of Tobin's q ratios are available from
the NBER manufacturing sector master file. (7) Firms should be included in the Investment Broker
Estimation System (IBES) database. The elasticity is measured by dividing percentage change of
postannouncement median earnings forecast compared to preannouncement median earnings
forecast by the percentage dividend change. The elasticity is set to 1 (-1) if it is greater (less)
than 1 (-1).

four groups have significantly positive mean elasticities. The one exception is the group of high-q firms that increased dividends. Note
that positive elasticity implies that the medians of analysts' earnings
forecasts change in the same direction as the dividend changes. The
most significant effect on cash flow expectations occurs for low-q
firms with dividend decreases. The mean elasticity is 0.56. Further,
out of 171 announcements, 145 have positive elasticity. For high-q
firms that decrease their dividend, the mean elasticity is 0.27. Out of
the 34 announcements, there were only 5 in which analysts revised
their forecasts in the direction opposite to the dividend change. Low-q
and high-q firms that increase their dividends have mean elasticities
of 0.08 and 0.01, respectively.13
According to the cash flow signaling hypothesis, the magnitudes of
signaling will be differentiated by the level of asymmetry in information between managers and shareholders. As shown in Table 1, low-q
firms tend to be smaller in size and have a larger dividend change.
13

Brous (1992) and O'Brien (1988) have found that analysts tend to be overly optimistic in their initial
annual earnings forecasts. Consequently, analysts' earnings forecasts are systematically lowered
each month up to the fiscal year end. The results reported in Table 4 do not control for this
optimism bias. To check for the extent to which an optimism bias may affect our results, we
follow the procedure discussed in Brous and Kini (1993) and find qualitatively similar results.

1012

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment Opportunities, and Dividend Announcements

Since asymmetry in information is greater for smaller firms and larger


dividend changes, the results that the magnitude of forecast revisions
is greater for low-q than high-q firms should be expected.
5.3 Revisions of long-term growth rate of earnings
In this section our goal is to determine the extent to which cash flow
signaling from dividends reflects future earnings as opposed to current
earnings. Analysts' forecasts of current earnings are dominated by the
forecasts of earnings from assets in place rather than from growth
options. To measure changes in the market's perceptions of growth
options we use analysts' forecasts of the long-term growth rate of
earnings.
Specifically, for each dividend announcement, we measure forecast
revisions from the previous month as the proportional change in the
analysts' forecasts of five year earnings growth. The abnormal forecast
revision of analysts' five year growth of future earnings is computed by
subtracting the average forecast revision estimated over the estimation
period (months -24 to -7 and months 7 to 24) from the forecast
revision for each month in the event period (months -6 to 6 relative to
the dividend month). The results are shown in Table 5 for months -4
to 4 relative to the dividend month. The results are divided by the sign
of the dividend change and by the level of investment opportunities.
For both low- and high-q firms, we find that analysts' forecasts
of five year earnings growth do not significantly change when firms
have a large increase in their dividends. Although this result could
be due to firms financing increased dividend payments through the
issuance of debt or equity, Long, Malitz, and Sefcik (1994) find that
firms do not issue debt to increase dividends. Thus, the results suggest that, consistent with Miller and Rock's (1985) signaling model,
dividend increases release managers' information about current cash
flows rather than future cash flows. This is also consistent with Lintner's (1956) finding that for healthy firms, the current earnings are the
major determinant of the dividend change.
We find a somewhat different result for dividend decreases. As
shown in Panel B, while the average abnormal forecast revisions are
also not significantly different from zero as measured by z-statistics,
the proportion of firms with positive abnormal forecast revisions for
the announcement month and for the following three months are all
significantly less than 50 percent. This result is basically driven by
low-q firms. (The results for high-q firms with dividend decreases
should be interpreted carefully as there are only 19 observations.)
Thus, unlike dividend increases, dividend decreases appear to signal
managers' views on both current and long-term cash flows.
Our finding that analysts lower their long-term earnings growth

1013

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

7he Review of Financial Studies / v 8 n 4 1995

2
*,

in

-24The
4 3 2 1 0 -1 -2 -3 -4 Panel
4 3 2 1 0 -1 -2 -3 -4 Panel
to
B:
A:
***) Wilcoxon
month
-7
analysts'
Forecast
abnormal
and
n
signed7
denotes
130129130131131127124122116Dividend
876879877880883853831827815Dividend
forecast
to
rank
of 24)forecast
(**,

Table
5
Monthly

average

test.

increases
decreases
from
significance
five-year
revisions
the
forecast
at
Average
0.0001
0.0314
0.0389
0.0172
-0.0038
-0.0087
0.0035
-0.0002
-0.0349
-0.0237
-0.0047
-0.0196
-0.0024
-0.0048
-0.0047 revisions
-0.0036
-0.0030
0.0058
abnormal
All
are
the
growth
10,
forecast
5
rate
0.51
0.51
0.51
0.47
0.52
0.43
0.49
0.54
0.52
0.50
and of computed0.39**
0.38**
0.40**0.48**2
0.36***
0.32***
0.40***
0.33***
1
positive
by
revisions
Proportion
future
for
percent

i11110i1111211210810510498
subtracting
months
level, earnings
-4
from
to average
the4.

267269268269270258248246242

abnormal

forecast

revisions
of
the

five-year

forecast
Average
0.0001
-0.0413
-0.0086
0.0463
0.0390
-0.0244
-0.0147
-0.0072
-0.0192
0.0222
0.0142
-0.0108
0.0194
-0.0083
-0.0107
-0.0149
0.0028
-0.0106 revisions
q< growth
abnormal
The
respectively.
1
forecast
of
previous
forecast
0.46
0.49
0.49
0.47
0.48
0.46
0.48
0.47
0.50
0.45*
revisions 0.40*
0.41**
0.35***
0.31***
0.36***
0.39***
0.32***
0.35***
month.
earnings
positive
revision
Test
is
estimated
19 19 19 19 19 19 19 18 18
statistics
from
defined
on to the
thebe

Proportion
609610609611613595583581573

n
share'

the

forecast
Average
-0.0172
-0.0015
-0.0123
-0.0195
-0.0213
0.0029
0.0603
0.0286
0.0042
0.0012
-0.0052
-0.0014
0.0040
-0.0007
-0.0012
-0.0022 revisions
-0.0040
-0.0006
q>
abnormal
estimation
1

proportion
period
proportional
0.51
0.42
0.32
0.42
0.47
0.42
0.37
0.33
0.33 0.48*
0.53
0.53
0.53
0.53 0.52
0.50
0.16*
0.57**
positive
arechange
(months

per

positive
Proportion

1014

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment Opportunities, and Dividend Announcements

forecasts following dividend decrease announcements contributes to


the existing literature concerning the type of information contained in
dividend announcements. DeAngelo, DeAngelo, and Skinner (1992)
analyze the dividend reduction decision of firms with current losses.
They conclude that dividend changes have some incremental information about future earnings over that conveyed by current earnings
in the sense of improving the ability of current earnings to predict
future earnings. Our result is complimentary to theirs and provides
direct evidence that dividend decrease announcements contain information on future earnings. This result is also consistent with Warther's
(1994) coarse dividend signaling model in which he predicts that dividends are more likely to have information when they are decreased
than when they are increased. In his model, only the "worst" firms
cut dividends, so that firms that do not cut their dividends reveal only
that they are not among the worst firms. This implies that dividend
reductions are more informative about future earnings prospects than
are dividend increases.
Our finding that only dividend decreases release information on
long-term earnings growth also provides an empirical explanation as
to why dividend decreases are associated with a larger absolute magnitude of stock price reaction than are dividend increases. For our
sample, the CAAR of -5.16 percent for the dividend decrease announcements is over four times as large (in absolute value) as the
CAARof 1.15 percent for the dividend increase announcements. This
differential stock price reaction between dividend increase and decrease announcements could be due to the problem of determining
the market's conditional expected dividend, since dividend decrease
firms have greater dividend yields and larger absolute values of dividend changes, and are smaller in size (see Table 1). However, the
differential still holds after controlling for these variables. We conclude
that the analyst forecast revision result suggests that the differential
reaction can be explained in part by information relevant to the longterm earnings growth rate.
6. Concluding

Comments

In this article we have provided tests of whether the information revealed by dividend change announcements is more consistent with
the cash flow signaling hypothesis or the Lang and Litzenberger (1989)
version of the free cash flow hypothesis. We find that the stock price
reaction to large (at least 10 percent) dividend change announcements
is generally consistent with the predictions of the cash flow signaling
hypothesis. Although we find that for dividend increases, the abnormal return for low-q firms is significantly larger than that of high-q

1015

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

The Review of Financial Studies / v 8 n 4 1995

firms, this differential reaction does not persist after controlling for
dividend change, dividend yield, and firm size.
Due to the relationships between the three control variables and the
investment opportunity set, we take an alternate approach to discriminate between the cash flow signaling and free cash flow hypotheses
as explanations of the wealth effects surrounding dividend change announcements. We directly examine the sources of the wealth effects
suggested by the two hypotheses. We find that dividend increase (decrease) firms experience significant increases (decreases) in capital
expenditures over the three years following the dividend change, a
result that is inconsistent with the implications of the free cash flow
hypothesis for dividend change announcements. We also provide significant evidence that announcements of dividend increases and decreases cause analysts to revise their current earnings forecasts in a
manner generally consistent with the cash flow signaling hypothesis.
In addition, we find that analysts tend to lower their long-term earnings growth forecasts following dividend decrease announcements,
but not following dividend increase announcements. This result potentially explains why dividend decreases cause a larger stock price
reaction than do dividend increases.
A central issue in any test of the free cash flow hypothesis is
the question of a measure for a firm's investment opportunities. Our
choices for proxies for the investment opportunity set were Tobin's
q ratio and the direction of insider trading. If our choices were poor
proxies for the investment opportunity set, our results on the market
reaction by the investment opportunity set are biased against the free
cash flow hypothesis. The examination of the sources of the valuation effects provides clearer evidence than the analysis of the wealth
effects, not only because the observed stock price reactions are affected by confounding factors, but also because our conclusion from
the examination of capital expenditure changes does not particularly
depend on how good a proxy Tobin's q ratio is. That is, we find significant changes in capital expenditures in the direction opposite to
the prediction by the free cash flow hypothesis for both high-q and
low-q firms, although the magnitude of the changes for high-q firms
are smaller than that of low-q firms.
Although our results indicate that the free cash flow hypothesis
does not explain the information effects of dividend change announcements, we cannot rule out the possibility that the free cash flow hypothesis explains the observed cross-sectional differences in dividend
policy. In particular, the fact that low-q firms have higher dividend
yield and larger dividend change than high-q firms is consistent with
the implications of the free cash flow hypothesis [Smith and Watts
(1992)].

1016

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

Signaling, Investment Opportunities, and Dividend Announcements

References
Aharony, J., and I. Swary, 1980, "Quarterly Dividend and Earnings Announcements
Stockholder's Returns: An Empirical Analysis," Journal of Finance, 35, 1-12.

and

Asquith, P., and D. Mullins, 1983, "The Impact of Initiating Dividend Payments on Shareholders'
Wealth," Journal of Business, 56, 77-96.
Bajaj, M., and A. Vijh, 1990, "Dividend Clienteles and the Information Content of Dividend
Changes," Journal of Financial Economics, 26, 193-219.
Bhattacharya, S., 1979, "Imperfect Information, Dividend Policy, and the 'Bird in the Hand'
Fallacy," BellJournal of Economics, 10, 259-270.
Brous, P., 1992, "Common Stock Offerings and Earnings Expectations: A Test of the Release of
Unfavorable Information," Journal of Finance, 47, 1517-1536.
Brous, P., and 0. Kini, 1993, "ARexamination of Analysts' Earnings Forecasts for Takeover Targets,"
Journal of Financial Economics, 33, 201-226.
DeAngelo, H., L. DeAngelo, and D. Skinner, 1992, "Dividends and Losses," Journal of Finance,
47, 1837-1863.
Denis, D., D. Denis, and A. Sarin, 1992, "The Information Content of Dividend Changes: Cash
Flow Signaling, Overinvestment, and Dividend Clienteles," working paper, Virginia Polytechnic
Institute and State University.
Eades, K., P. Hess, and E. Kim, 1985, "Market Rationality and Dividend Announcements," Journal
of Financial Economics, 15, 3-34.
Easterbrook, F. H., 1984, "Two Agency-Cost Explanations of Dividends," American Economic
Review, 74, 650-659.
Eddy, A., and B. Seifert, 1988, "Firm Size and Dividend Announcements," Journal of Financial
Research, 11, 295-302.
Fehrs, D., G. Benesh, and D. Peterson, 1988, "Evidence of a Relation between Stock Price Reactions
around Cash Dividend Changes and Yields," Journal of Financial Research, 11, 111-123.
Healy, P., and K. Palepu, 1988, "Earnings Information Conveyed by Dividend Initiations and
Omissions," Journal of Financial Economics, 21, 149-176.
Howe, K., J. He, and W. Kao, 1992, "One-time Cash Flow Announcements and Free Cash-Flow
Theory: Share Repurchase and Special Dividends," Journal of Finance, 47, 1963-1975.
Jensen, M., 1986, "Agency Costs of Free Cash Flow, Corporate Finance, and the Market for
Takeovers," American Economic Review, 76, 323-329.
John, K., and L. Lang, 1991, "Insider Trading around Dividend Announcements:
Evidence," Journal of Finance, 46, 1361-1389.

Theory and

John, K., andJ. Williams, 1985, "Dividends, Dilution and Taxes: A Signaling Equilibrium," Journal
of Finance, 40, 1053-1070.
Kalay, A., 1980, "Signaling, Information Content, and the Reluctance to Cut Dividends," Journal
of Financial and Quantitative Analysis, 15, 855-869.
Kalay, A., and U. Loewenstein, 1985, "Predictable Events and Excess Returns: The Case of Dividend
Announcements," Journal of Financial Economics, 14, 423-449.
Kalay, A., and U. Loewenstein, 1986, "The Information Content of the Timing of Dividend
Announcements," Journal of Financial Economics, 16, 373-388.

1017

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions

The Review of Financial Studies / v 8 n 4 1995

Keown, A., P. Laux, andJ. Martin, 1992, "JointVentures, Latent Assets and the Information Content
of Corporate Announcements," working paper, Virginia Polytechnic Institute and State University
and University of Texas at Austin.
Lang, L., and R. Litzenberger, 1989, "Dividend Announcements: Cash Flow Signaling vs. Free Cash
Flow Hypothesis?," Journal of Financial Economics, 24, 181-191.
Lang, L., R. Stulz, and R. Walkling, 1991, "A Test of the Free Cash Flow Hypothesis: The Case of
Bidder Returns," Journal of Financial Economics, 29, 315-335.
Lehn, K., and A. Poulsen, 1989, "Free Cash Flow and Stockholder Gains in Going Private
Transactions," Journal of Finance, 44, 771-789.
Lindenberg, E., and S. Ross, 1981, "Tobin's Q Ratio and Industrial Organization," Journal of
Business, 54, 1-32.
Lintner, J., 1956, "Distribution of Incomes of Corporations among Dividends, Retained Earnings
and Taxes," American Economic Review, 46, 97-113.
Long, M., I. Malitz, and S. Sefcik, 1994, "An Empirical Examination of Dividend Policy Following
Debt Issues," Journal of Financial and Quantitative Analysis, 29, 131-144.
Miller, M. H., and K. Rock, 1985, "Dividend Policy under Asymmetric Information," Journal of
Finance, 40, 1031-1051.
O'Brien, P. C., 1988, "Analysts' Forecasts as Earnings Expectations," Journal of Accounting and
Economics, 10,187-221.
Ofer, A., and D. Siegel, 1987, "Corporate Financial Policy, Information, and Market Expectations:
An Empirical Investigation of Dividends," Journal of Finance, 42, 889-911.
Perfect, S., D. Peterson, and P. Peterson, 1994, "Self-Tender Offers: Differentiating the Effects of
Free Cash Flow and Cash Flow Signaling," forthcoming in Journal of Banking and Finance.
Perfect, S., and K. Wiles, 1994, "Alternative Constructions of Tobin's Q: An Empirical Comparison,"
Journal of Empirical Finance, 1, 313-341.
Pettit, R., 1972, "Dividend Announcements, Security Performance, and Capital Market Efficiency,"
Journal of Finance, 28, 993-1007.
Pilotte, E., 1992, "Growth Opportunities and the Stock Price Response to New Financing," Journal
of Business, 65, 371-394.
Rozeff, M. S., 1982, "Growth, Beta, and Agency Costs as Determinants of Dividend Payout Ratios,"
Journal of Financial Research, 5, 249-259.
Servaes, H., 1991, "Tobin's Q and the Gains from Takeovers," Journal of Finance, 46, 409-419.
Servaes, H., 1994, "Do Takeover Targets Overinvest?," Review of Financial Studies, 7, 253-277.
Smith, C., and R. Watts, 1992, "The Investment Opportunity Set and Corporate Financing, Dividend,
and Compensation Policies," Journal of Financial Economics, 32, 263-292.
Warther, V., 1994, "Dividend Smoothing: A Sleeping Dogs Explanation," working paper, University
of Southern California.

1018

This content downloaded from 111.68.99.250 on Fri, 5 Dec 2014 08:36:55 AM


All use subject to JSTOR Terms and Conditions